[WIP] Hardcoded Subtitles to SRT

An all-in-one tool for video subbing and dubbing. Done by creating an SRT file from a given video file that has hardcoded subtitles, finally generating an audio dub on the extracted subtitle.

The scripts relies on Tesseract for the optical character recognition, and FFmpeg for parsing and converting audio/video files.

Features (TODOs)

• Parse SRT-file (either in another language / same language) and extract hardcoded subtitles in given timestamps (Fast)
• Search for and extract the hardcoded subtitles by going through the frames and calculating the timestamps (Slower)
• Web-based GUI editor for easier configuration and semi-automatic corrections
  • Configure total frames to extract from
  • Configure the subtitle crop-box position by specific frames
  • LLM API used for sentence correction
• Process SRT-file and generate audio dubbing through a given TTS model
• All in one tool for easier video subbing and dubbing

Useful commands

To convert mkv to mp4 run:

• ffmpeg -i video.mkv -codec copy video.mp4

To convert mp3 to wav run:

• ffmpeg -i audio.mp3 -acodec pcm_u8 -ar 22050 audio.wav

To convert mp4 to mkv run:

• ffmpeg -i input.mp4 -vcodec copy -acodec copy movie.mkv

To extract sub channels from a video file run:

• ffmpeg -i movie.mkv -map 0:s:0 subs.srt

Standalone scripts

To extract hardcoded subtitles from a video file run:

• python3 ./srt_extract/hard_subs_to_srt.py movie.mkv output.srt helper.srt

To process and cleanup srt files run (outputs subbed.srt + filtered.srt):

• python3 ./tts_srt_parsing/process_tts.py input.srt

To generate tts from srt files run (outputs small tts chunks):

• python3 ./tts_srt_parsing/tts-edge.py input.srt

To process the generated tts into an audio file ():

• python3 ./process_video/standalone.py

Note that the script currently only support HD (1280x720) videos for now

How to use

TBA

Dependencies

• Python 3
• Tesseract OCR with support for your target language
• FFmpeg to convert video to a supported format

To download the required python package dependencies, run

pip3 install -r requirements.txt

Development

TBA

Docker

TBA

Credits

The codebase builds upon the foundations of the following projects:

• victorjoh for the initial code that this project were based on
• rany2 for the Microsoft Edge's tts python wrapper
• vsakkas used as inspiration in reverse engineering the Bing Chat / Copilot / Sydney API
• video-subtitle-extractor for the VideoSubFinderCli binary 🙏